My program at MIT is a short two years, and I find myself now at the halfway point. Which is a bit disconcerting, honestly, as
- I am absolutely loving my time here and do not want it to end any time soon, and
- at some point in the very near future I will need to figure out exactly what it is I’m doing here, and then turn that into a thesis project.
So, as this glorious northeast summer passes the halfway point and September begins to loom on the horizon, I sit down to organize my thoughts.
Admittedly, the semester started out rough. Marvin Minsky passed away on January 24th, just a week before classes resumed. His death was a somber beginning to the semester, and his absence from the campus was felt (and discussed) for months, in nearly every class. More personally, I developed pneumonia in January, which stayed with me all the way into April. Luckily, MIT has an extended winter break that lasts all of January, and so I was able to crash on the couch and recover with the help of copious amounts of chicken noodle soup and MST3K before the spring semester started. Even so, I still felt entirely exhausted when it came time to select the semester’s classes. Psychologically and physically, I was not game for the slightly-overloaded schedule I had entertained in the fall. And I so opted to take fewer hours. In retrospect, this hardly changed much, especially as I also agreed to TA two classes during the semester. I guess I’m not very good at “taking it easy” – maybe because the program is so short, and I recognize the need to make best use of my brief time here.
My classes once again gravitated towards AI. The most interesting class of the semester was The Human Intelligence Enterprise, taught by Patrick Winston, my AI professor from the fall. The class was a more historical/theoretical review of AI and largely consisted of reading/writing assignments on various seminal AI papers from the last 50-60 years. I also took Human-Machine Symbiosis, a Media Lab class taught by Pattie Maes, whose name I’d known as a pioneer of human-computer interaction research for years. Unfortunately that class felt a bit rushed – the majority of the time was spent covering various projects in Maes’ Fluid Interfaces group, and not much was given to the critical discussion I had hoped for. The computation first years also all took a Pre-Thesis Prep (half) class, which, while interesting, probably did more to bring us together as a friend group (for our weekly Computation Lunches) than anything. Finally, I attended Cognitive Science as a listener – a fascinating class co-taught by Pawan Sinha, Josh Tenenbaum, and Ted Gibson.
Early in January, I started work for Neri Oxman as a TA for her class Design Across Scales. The class brought together a survey of various professionals speaking about design in a number of fields, alongside a series of labs designed to teach the students basic design skills. I really enjoyed it – it made for a nice break from the endless reading and writing, and the lecturers were absolutely stellar. I also TA’d Sheila Kennedy’s Paper Light Workshop, which explored ways to reconsider the design and implementation of street lights in impoverished Indian communities. The class was very small, so it was a pretty fluid role – admin roles, student participation, some teaching opportunities. This work obviously doesn’t closely tie in to my primary interests, but I enjoyed the opportunity to engage in discussions about design, as well as interact with these prominent figures in the field – in fact, it’s thanks to Neri that I now have a summer job at Moshe Safdie’s office!
I think I can say that the first year of grad school has been about breadth, while perhaps the second should be about depth. I have been taking classes from other subjects (computer science, neuroscience/cognitive science, etc.), searching for possible ways to connect them to design. I guess in some ways I’m trying to learn what mistakes NOT to make – if I am going to connect disparate subjects, then I certainly need to know how to talk about them. By researching what has and hasn’t been done in the fields, I can hopefully find the right points to drill down with my own research. This has also posed a difficulty, however: Throughout the semester I repeatedly came to the “edge of the shallows” in several areas. I’ve surveyed these topics well, but in order to proceed further, I feel like I needed extensive study (or even an additional degree).
And so, what will be my depth? I did begin to find resonance in some topics that might shed light on this, especially towards the end of the semester. In a short reading of Parts IV and V of The Emotion Machine, I was overjoyed to read about Minsky’s Model Six approach to cognitive activity, which posits that there are six levels of thought – instinct, learned reaction, deliberation, reflection, self-reflection, and self-reflective emotion – and, more interestingly, that these levels must be interconnected and easily switchable in order to display behaviors such as imagination and prediction. I also became very interested in theories of empathy and analogy as flexible structures for navigating complex, previously uncharted territory (Gentner and Markman, McIntyre, perhaps even Minsky again with “K-lines,” etc.). I think these topics interested me because they revealed a looser, less prescriptive approach to consciousness: One where we might “feel” our way to understanding by making connections with different, perhaps far simpler situations.
In my final project for Human Intelligence Enterprise, I studied the topic of iteration as it’s used in design. Specifically, I looked at iteration as a possible means for creative emergence (as opposed to combinatoric emergence – a distinction made both by Cariani and Boden). I do find this to be an important topic that may require further study this semester: Can emergence be modeled computationally? Is it used by designers? If so, is it something we control, or is it an “irrational” by-product that we just make use of? Is ignorance (either inherent or imposed) required for design, and if so, how then do we control it?
This brings up another topic of interest – an enduring one, that I hope will find its way into my final thesis: alternative / weird / irrational knowledge. At the end of the day, as much as I enjoy the fruits of computation, I am more interested in what makes us different – “What Human Can’t Do,” to adapt Dreyfus’ famous phrase. I am interested in defining us this way – not as a fearful retreat from computation, but as a way to understand how we relate to it, and it to us. Another hot-button topic of discussion during the semester was AlphaGo’s defeat of world Go champion Lee Sedol. While there’s no denying that this was an incredible achievement for the world of computation, I found myself wondering: Isn’t it more interesting that we are able to play (and master) the game of Go at all?
I’d like to challenge the premise that we are inherently (only) rational beings, and open up a discussion about what it means to embrace irrational behavior. For decades we have sought to develop computers to do exactly what we want them to – what happens when we loosen our grip a bit?
But I shouldn’t go further, at the moment, as I’ve only just started to scratch at these topics after a 7+ week hiatus, and my thoughts are a bit rusty. Needless to say, this degree is going by fast – I suppose I had some notions that grad school would be a time for me to develop a skill set AND use it, but that increasingly seems like a misconception. I don’t really have time to get burnt out and then recover, so I’m welcoming the more relaxing summer working in Safdie’s office. The weather has been glorious and the opportunities for travel, exploration, and fun have been abundant – hence the tardiness of the post…In fact, today begins my series of Tuesday “study nights,” whereby I attempt to scrape at least a tiny bit of scholarly productivity out of the semester. I highly doubt I’ll accomplish the reading, writing, and thought-organization that I had optimistically hoped for as the summer started, but even so, I won’t be disappointed: Sometimes you just need to let thoughts cook on their own a little bit before they really become clear. Explain THAT in computational terms.
Scott Penman 